OCTIS

Why we included this project

Topic modeling in Python usually means stitching together separate libraries for each stage of the pipeline. OCTIS wraps that workflow into one package, covering classical models like LDA and NMF as well as neural topic models, and it adds Bayesian optimization to search hyperparameters against a chosen evaluation metric automatically. The package was accepted at the EACL 2021 demo track and includes preprocessed benchmark datasets and standard coherence and quality metrics, so comparisons between models are fair and reproducible. A web dashboard lets you launch and monitor optimization runs without writing code, while the Python API covers the same workflow for scripted pipelines. Researchers and engineers who need dependable, side-by-side topic model comparisons will find it a solid fit.